AI in action · Markets
Ives: Up to 20% of Planned Data Centres May Not Be Built
Analyst Dan Ives estimated 10% to 20% of planned US data centres may not materialise after the midterms, though he says demand still outruns supply four to one.
The week that ended on Friday, November 13, rearranged the AI trade more than it shrank it. The Nasdaq had its best week since August even as Nvidia, the sector's anchor, fell for a sixth week in a row to $162.85 after Jensen Huang resigned as chief executive. CNBC's weekend review of the trade looks at two forces now acting on it at once: a leaderless Nvidia and a Congress that changed hands on November 3.
On the first, long-time bull Dan Ives cut his Nvidia price target to $225 from $300 while keeping a buy rating. He wrote that the company is "roughly 2-3 years ahead of the nearest competitor" and that its valuation is too low once the turmoil passes. Interim chief executive Colette Kress faces her first market test with Tuesday's results.
The second force is political. Democrats won control of Congress, and voters in several states backed restrictions on new data centres. Ives estimated that 10% to 20% of planned data centres may not materialise, while arguing that demand for AI computing still exceeds supply by about four to one. Dave Mazza, chief executive of ETF issuer Roundhill, linked the election result to a possible slowdown in the build-out.
Fund flows showed where money went. Roundhill's Magnificent Seven ETF gained less than 1% on the week; Mazza argued that the largest technology companies could benefit as investors look beyond Nvidia. Its neocloud ETF, which holds the smaller companies renting out AI computing power, rose just over 1% but is down 39% since its launch in August. CoreWeave had its best day since April 2025 after results on Wednesday, then fell again on Friday, and Nebius dropped after its own report. AMD rose nearly 13% on the week, with Broadcom close behind, before both gave back part of it on Friday.
The pattern matters more than any single price target. Investors are separating the AI trade into layers, from cash-rich hyperscalers to leveraged builders and chipmakers, and pricing each one for its own exposure to delays. If a fifth or so of planned capacity is never built, the companies with the strongest balance sheets lose the least, and the ones that borrowed to build lose the most.
Sources: CNBC — Nvidia AI trade: Huang exit, Congress reshape Big Tech stock outlook (Nov 14, 2026)